i want to use cuda in rust natively ,but the library in github have bug or have been deprecated
What do you mean with "native" support? Having cuda support as part of the standard library (unlikely to ever happen. there is no benefit over having an external library) Compiling rust code to nvptx bytecode? (There is GitHub - Rust-GPU/Rust-CUDA: Ecosystem of libraries and tools for writing and executing fast GPU code fully in Rust., but it hasn't been updated in two years) Or do you mean something else?
3 Likes
My workflow for this is to write your GPU kernels in CUDA C++, compile them to a fatbin with nvcc, and use the cust crate to dispatch them, allocate memory, etc.
1 Like
Is Rust-CUDA actively maintained ? GitHub - Rust-GPU/Rust-CUDA: Ecosystem of libraries and tools for writing and executing fast GPU code fully in Rust. I think what the op is asking requires 1st class support from NVidia for Rust alongside C/C++ in the nvcc compiler. Not sure how the CUDA unified memory model could conceptually be supported by the borrow-checker ... ZLUDA GitHub - vosen/ZLUDA: CUDA on AMD GPUs was a CUDA API written in Rust (targetting AMD GPUs) - but AFAIK NVidia got their lawyers to update the EULA to ban transcompilation layers Nvidia bans using translation layers for CUDA software — previously the prohibition was only listed in the online EULA, now included in installed files [Updated] | Tom's Hardware
2 Likes
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.